Various mapping applications can be accessed by way of the Internet. Many of these mapping applications allow a user to review aerial images of a particular geographic region. Thus, if desired, a user can zoom in on a particular geographic region to obtain some knowledge regarding, for instance, roadways near an intended destination, surrounding buildings, amongst other information. Moreover, mapping applications can be used to provide to a user an indication of where certain types of businesses are located such as restaurants, shopping centers, etc. Furthermore, many mapping applications can provide the user with driving directions between a first location and a second location, and can provide the user with visual information pertaining to the route between the first location and the second location.
Some of these mapping applications are being adapted to provide even more granular information. For instance, some mapping applications are being adapted to include photographs of geographic regions such as particular cities from street-level. Thus, a user can provide a particular location (e.g., an address) to the mapping application, and the mapping application can be configured to provide the user with a three-dimensional representation of buildings pertaining to the address. Such a three-dimensional representation can include images of facades of buildings, for example, wherein the images are obtained by mounting a digital camera on an automobile and capturing images as the automobile traverses through the city.
To allow for automatic three-dimensional reconstruction from captured images, a minimum forward overlap between images is required. For instance, such minimum overlap may be between 20 and 90 percent. In another example, the minimum overlap may be between 60 and 80 percent. To ensure such overlap, conventional techniques for obtaining images have included taking images at a defined, relatively high frequency. While this approach can ensure the requisite overlap, it may result in obtaining a substantially greater amount of data than is needed to generate the three-dimensional representation of a building, street, etc. Such a substantial amount of data can render processing of such data tenuous and computationally intensive. Another conventional mechanism for ensuring the requisite overlap is using a GPS sensor to cause the digital camera to obtain images at predefined distance intervals. More particularly, the digital camera is triggered based on a position measurement from a global positioning system. This approach can fail, however, where GPS reception is suboptimal, which can often occur in congested cities due to buildings blocking GPS satellites.